Cindy is an open-source AI agent that works out of the box. She brings multiple harnesses, models and tools into one agent that finishes real work in your projects and apps. Ready from day one. Yours to shape over time.
Cindy starts with Claude Code and Codex, with more harnesses on the way and a Cindy-native harness brewing. Every supported model runs across every supported harness. Switch either one mid-task while your workspace, memory, skills and tools stay continuous.
Take the best from every stack. Get trapped by none.
Switch harnesses or models mid-task. The work never starts over.
A task does not have to bet on one combination from start to finish. CLAUDE CODE × FABLE 5 plans first, three workers—GROK 4.5、KIMI K3、GLM 5.2—execute in parallel, and CODEX × GPT-5.6 performs an independent review. Harnesses and models pass the baton while Cindy holds the workspace, context and acceptance criteria together.
Architecture · Work breakdown · Acceptance criteria
Diff review · Edge cases · Pass / Rework
Use Cindy's managed service, a Coding Plan, your own API key or local models. All four connect directly, with 20+ tools and sensible defaults ready to go. Download, sign in and start working.
We encourage you to use the subscriptions and keys you already own.Authorize a Claude Code or Codex Coding Plan once and keep using it inside Cindy—no duplicate bill.
Reuse existing environments and signed-in browser sessions with progress always visible.
GitLab, Confluence and Google Workspace connect too.
Move cleanly across local, remote and mobile environments.
Add tools and turn working methods into reusable capabilities.
Delegation starts with control. She knows when to stop, and you always have the switch. You do not have to trust a promise—every item below is yours to verify.
Judgment returns to you at critical moments. She will not make irreversible decisions on her own.
Code and file changes start in an isolated workspace instead of overwriting the original.
Every edit enters the review panel for line-by-line approval.
Return to an existing checkpoint and explore another approach when a result is not right.
Files, code and chat history stay on your computer by default; Cindy does not upload the whole project to its cloud.
See what she is doing, where she is and what it costs in real time.
Cindy works on your computer while control stays in your hand—even across remote servers and private networks.
Check progress, ask follow-ups and approve sensitive actions from mobile.
@ her in Slack or Feishu. The task runs on your computer and the result returns to the thread.
Hand off recurring work completely. She runs it on time and sends the result to your IM.
Changing harnesses, choosing models and connecting tools is how you dress her up. Teaching rules, skills and schedules is how you train her. Go deeper through MCP, plugins or the source itself.
Open source means more than visible. It means changeable.Apache-2.0 · Audit · Fork · Build · Contribute
Correct her once; every harness gets it right next time.
LIVETeach once, reuse forever, and share it with a team.
LIVEShe schedules, runs and reports recurring jobs herself.
LIVEConnect internal tools and systems so she grows into your environment.
LIVEChange features, UI and interactions, then share them through an open marketplace.
BUILDINGAudit, fork and build; contribute improvements and shape what Cindy becomes.
Apache-2.0Proactive briefs and work suggestions are in development—so she can spot useful work instead of only waiting for instructions.
Ready out of the box. Never boxed in.
Advanced teams already work alongside AI
Traditional collaboration is a pipeline: product waits for design, design waits for engineering, engineering waits for QA. Give everyone Cindy and put her in the team chat: ideas become discussable drafts on the spot, discussions become active tasks, and the pipeline becomes a network that starts together.
Turn an idea into a clickable demo during the meeting and discuss the artifact, not the spec.
Apply a direction to the real page so the team sees the result, not another mockup.
Run tests, docs and release notes in parallel, then focus on review and judgment.
@ Cindy in Slack or Feishu and the first working version can be there before the meeting ends.
Individuals should not start with setup, and teams should not repeat it for every person. Configure models, access, credits, skills and policies once; members inherit them when they sign in.
Cindy itself is free. Bring your own key or existing subscription and use every software capability without paying Cindy extra. The managed model service works on sign-in, deducts usage transparently and saves 20% on a monthly plan.
See full pricing and rules →
